Scholar
Arthur Jacot
Google Scholar ID: G6OhFawAAAAJ
Assistant Professor, Courant Institute of Mathematical Sciences, NYU
Citations & Impact
Citations
5,478
H-index
14
i10-index
14
Publications
20
Co-authors
13
list available
Contact
Publications
Resume (English only)
Academic Achievements
- Published multiple papers at top-tier conferences including ICLR, NeurIPS, and ICML
- 2025: 'Wide Neural Networks Trained with Weight Decay Provably Exhibit Neural Collapse' (ICLR oral)
- 2025: 'Hamiltonian Mechanics of Feature Learning: Bottleneck Structure in Leaky ResNets' (CPAL oral)
- 2023: 'Implicit Bias of Large Depth Networks: a Notion of Rank for Nonlinear Functions' (ICLR spotlight)
- Research focuses on feature learning, implicit bias, geometry of loss landscapes, and training dynamics in deep neural networks
Background
- Assistant Professor at the Courant Institute of Mathematical Sciences, NYU
- Aims to develop new mathematical concepts and tools to describe the training dynamics of Deep Neural Networks
- Seeks to build a Theory of Deep Learning that transforms how AI models are trained and developed
- Currently most excited about showing that DNNs implement a computational version of Occam's razor—finding the fastest algorithm/circuit fitting the training data
- Also interested in feature learning, emergence of low-dimensional representations under weight decay, and identifying different training regimes (e.g., NTK regime, active regime)
